After much hard work, Yamaha and Steinberg persuaded Rupert Neve that they could faithfully model his latest hardware in plug-in form. We pit the virtual Porticos against the real things...
Last year's announcement of the partnership between Yamaha and Rupert Neve Designs probably raised an eyebrow or two, considering the differences in company size and product portfolio. However, when the R&D department at Yamaha — led by Toshifumi 'Dr K' Kunimoto — managed to convince Mr Neve that they could faithfully replicate his hardware in the digital domain, he agreed to be a part of the project, sharing his knowledge and passion for analogue circuit design. After two years in development, using Yamaha's Virtual Circuitry Modelling technology, the plug-ins finally earned the approval of Mr Neve. Dr Kunimoto was kind enough to share some insights into the process, as collected in the 'Interview With Dr K' box.
Some of Rupert Neve's designs have been around for more than 60 years, and many of them hold iconic status. Rupert Neve Designs Inc, on the other hand, have been around since 2005, and manufacture the Portico series of hardware units which — much like their predecessors — are designed like modules in a mixing console. They have yet to earn the almost religious status of the old Neve modules, but still represent a mighty fine range of outboard with impressive audio quality. At the respectable age of 85, Mr Neve is still a force to be reckoned with.
The Portico 5033 EQ is a five-band EQ containing three fully parametric bands with variable Q control, flanked by low and high shelving filters. The shelves have a frequency range of 30Hz to 300Hz and 2.5kHz to 25kHz respectively, and share a bypass button. The parametric filters have individual bypass, a filter width (Q) of 0.7 to 5, and overlapping frequency ranges between 50-400Hz, 330-2500Hz and 1.8-26kHz. All five filters and the input trim knob have a gain range of ±12dB. The plug-in version looks just like the hardware's front panel except for an added graph displaying the EQ's response curve. Like the hardware EQ, the plug-in lacks high- and low-pass filters.
The Portico 5043 compressor features the usual compression parameters, with a threshold control that runs from -50dB to 0dB, ratio from 1.1:1 to limit, attack time from 20ms to 75ms, release 100ms to 2.5s and a make-up gain range from -6dB to +20dB. The LED-style input meter keeps track of the input level, and the gain-reduction meter indicates the level of compression. A more unusual feature is the Feedback switch, which changes the compression topology from feed-forward to feedback detection. In feed-forward mode, the compressor reacts to the input level and responds very fast and accurately to changes in the dynamics. Switching to feedback mode moves the level detection to the output signal; this alters the compression characteristics, making it less responsive to changes in dynamics and prone to 'overshooting' transients. The result, however, is often perceived as more musical and warmer-sounding, and almost all of Rupert Neve's earlier compressors were of feedback circuitry design. The option of both old and new styles of compression makes the 5043 very versatile.
As a proud owner of two Neve Portico 5033 EQs, I am very accustomed to their sound, which is not directly comparable with vintage Neve outboard. The classic Neve EQs have a tendency to inflate the lows and low-mid in a very charming way, whereas the Portico EQ sounds much more transparent, adding a low end that is (just) firm and controlled. What I love most about the Portico EQ is the way it can add treble without ever sounding harsh — it's silky-smooth even at high gain — and its ability to lift mid-range without diminishing anything else in the frequency spectrum.
It was with great anticipation that I opened the Steinberg plug-in emulation for a test spin. Bass and treble are usually the most critical areas for an EQ to process, so I began by testing the shelving filters on a couple of different sound sources. Turning the knobs, I almost instantly recognised the behaviour of my hardware units: the low end is exactly as firm and controlled as the hardware and the treble is open and smooth-sounding. When I activated the parametric filters and swept from the low-mid up to the high end, it felt just like tweaking the hardware, because the plug-in can push one frequency range without diminishing another. It's sometimes hard to describe but, to me, digital filters have a tendency to affect more than just the frequency ranges they amplify, in a way that high-quality analogue filters do not.
In order to check whether the plug-in really measured up, I filtered a couple of instruments using the hardware EQ until they sounded as good as possible. I then measured the filtered frequency response with test tones and tweaked the plug-in until the frequency response was as close as possible, with an accuracy of about 0.1dB to 0.2dB throughout the spectrum. When I compared the sound clips the next day, performing a blind test, I couldn't tell them apart, so my conclusion has to be that the EQ plug-in sounds just as good as the original hardware — which is very impressive!
I've always felt that just running audio through the hardware EQs makes something nice happen, even in bypass mode, and I get the same feeling by just activating the Steinberg plug-in. Maybe it's the fact that both start rolling off the low end at about 30Hz, taking away possible low-end mud, but it's more likely that the transformer-coupled inputs and outputs add something that's just too subtle to measure.
A colleague was kind enough to lend me his Portico 5043 compressor to test against the Steinberg emulation, and I spent a full day compressing instruments and mixes with it. Interestingly enough, it's been criticised on Internet forums by people comparing it to vintage Neve compressors, and now I understand why, because the Portico compressor is super-transparent compared to its older cousins. After a bit of testing, though, I realised that the criticism is completely unfounded, because even though the compression is transparent, it's still musical and can even add some bite. It has a kind of openness and softness, yet is capable of compressing pretty heavily and even pumping. Compared with a FET compressor like the classic Urei 1176, it doesn't add any noticeable harmonic distortion or graininess. This makes it perfect for compressing instruments and mixes almost without being noticed. Not surprisingly, it works very well for drums that are in need of a bit of compression without changing the overall sound. Same goes for acoustic guitar.
When I compared the plug-in with the hardware it was, again, difficult to distinguish them. After a rigorous calibration with test tones at different signal levels, I processed a bunch of sound sources for later comparison. The sound clips demonstrated how well the plug-in mimics the performance of the hardware; wearing headphones, I could barely distinguish some of the clips, and listening even more closely, I felt that in the cases where I could do so, it was probably because I hadn't set the exact same amount of compression. At the fastest attack setting of 20ms, it looks and sounds as though the hardware unit responds a tad faster than the plug-in. Neither the plug-in or the hardware unit's gain-reduction meter shows the correct amount of gain reduction, and the readings also differed between the two. Even though I calibrated the compression by measuring the signal drop at different input levels, it was impossible to dial in the exact same attack and release settings. Nevertheless it was interesting that when I tried to pick the best-sounding clip in a subjective blind test, in the first round I picked the plug-in and in the second round, the hardware. Therefore the conclusion must be that the plug-in sounds just as good as the hardware — which, again, is very impressive.
During the test period, I used both plug-ins extensively in my productions, and the 5033 is more or less the only EQ plug-in I use nowadays, partly because I'm so used to the hardware, but mostly because it sounds better than other plug-ins. Like the hardware, it's very musical and just works in most situations. The only things missing are low- and high-pass filters, which is a pity.
The 5043 compressor has become my new favourite drum bus compressor, and along with UAD's Fatso emulation, it covers the entire range from clean to rather coloured compression. It also works really well for parallel drum compression, just adding weight without colouring the sound. Due to its ability to be transparent without sounding boring, it also works really well on acoustic guitar, controlling the dynamics without 'putting the lid on', and I was pleasantly surprised when I tried compressing a dynamic tenor voice by about 6dB: the results made it more consistent without sounding compressed. The 5043 plug-in is also a pretty decent mix-bus compressor with ratio set between 1.5:1 and 2:1 and the gain reduction never showing more than 2-3dB, but personally I still prefer an analogue compressor for mix-bus duties. Just like the hardware, the plug-in lacks an auto-release mode, so it's not entirely suited to acting as a mix-bus compressor.
Given the sheer sound quality of the plug-ins, one might presume that they are CPU hogs, but in fact this is not the case at all, and my Core2 Quad 2.4GHz CPU handled 42 stereo tracks with a 5033 and a 5043 on each channel without a glitch.
Steinberg and Yamaha have successfully manage to recreate two modern classics, under the guidance and with the blessing of Mr Rupert Neve himself. The sound quality is exactly as good as the hardware, which contributes a price tag well above that of most plug-ins. For that kind of money it's possible to buy a whole bunch of other EQs and compressors, but they won't give you the exact same sound as the RND Portico plug-ins and hardware. Demo versions are available at the Steinberg web site in VST3, VST 2.4 and AU versions, and all you need is an eLicenser dongle and a compatible DAW host to hear if they are worth the money. My conclusion is that the EQ is top-notch and very musical, and the compressor has its own personality, sounding transparent but with a bite. It will be interesting to see what Steinberg and Yamaha have in store for the future.
In the interests of thoroughness, I exerted myself by comparing the Steinberg RND Portico 5033 EQ with a whole bunch of different EQ plug-ins. After two hours, I'd managed to find only two other plug-ins with similar qualities, namely Flux's Epure II and Sonalksis' SV517 MkII. However, neither of them could create the exact same kind of shelving filters, because Epure has wider curves and Sonalksis has a different type of curves.
Digital compressors are pretty good at transparent compression, so initially I thought it was going to be an easy task finding an equivalent to the 5043 plug-in. But after scrutinising a number of different plug-ins, I came to the conclusion that the 5043 actually has its own identity. The difference lies mainly in the ballistics and the recovery ramp (release), which sounds more open and vibrant compared to its competitors. The attack also feels more authentic and analogue, if you like, with other plug-ins comparatively creating a more fluttering and uneven compression.
If you want to listen for yourself and see if the plug-ins measure up to the performance of the hardware, my comparative sound clips can be found on the SOS web site at /sos/dec11/articles/rndmedia.htm.
How did you convince Mr Rupert Neve that you could faithfully replicate his hardware?
While developing audio effects for digital consoles, we came up with new technologies that ensured a remarkable sound quality for emulating vintage studio gear and guitar stompers. I introduced him to our modelling technology through several examples. I also explained to him how his equipment is great in a mathematical sense. It may be too theoretical for the readers of SOS magazine, but there are many brilliant aspects regarding the transfer response of the equipment that Rupert has designed. I had been making many investigations to understand this kind of vintage gear, and therefore I know the secret that his gear holds — in a circuit theoretical sense. Rupert also realised that we go to great lengths to achieve a top sound quality for all our equipment. In the end, this is what convinced Rupert that Yamaha would be able to replicate his gear in the digital domain exactly.
How did Rupert Neve himself contribute to the project?
We met many times during the past four years, and talked a lot in detail about his hardware, technologies and sounds — not only from a circuit-theoretical point of view, but also looking at aspects of musical timbre. Most of the time Rupert's approach was very clear to us, as we're engineers who understand how analogue gear is designed, but at times it got very philosophical and was therefore more difficult to comprehend, especially in the area of tonal quality. And although Rupert is at home in the analogue world and we primarily live in the digital domain, fundamental ideas are described in the same 'language', allowing our discussions about realising replicas of his technology in the digital domain to continue very smoothly, the fruitful results of which were then implemented digitally.
What were the biggest technical challenges, and how did you overcome them?
There were two challenges: efficiently realising feedback-type compression in the digital domain, and achieving the approximations of the many transfer response(s) that the analogue EQ hardware has. We were very lucky to have resolved these essential issues before approaching Rupert, as this was the main reason for Rupert Neve Designs to choose Yamaha, keeping in mind that they were also approached by other companies.
This moment came to us very gradually. When we revealed the first prototype to RND, Rupert smiled and told us that we had an accurate model that virtually sounded like the original hardware. However, we were happy to know that he still required several weeks of measurements and serious listening tests. He then emailed us some points on subjective listening tests, and we continued to tune the model. After tweaking the model according to the many e-mails and face-to-face conversations that we had, RND sent us the 'approval certificate' for digitally modelling the Portico hardware. We spent approximately two years on the plug-ins. As we're situated in Hamamatsu, Japan, and Rupert Neve Designs are based in Wimberley, Texas, the long-distance collaboration wasn't exactly easy.
Do you simulate the circuit designs down to component level — for example, simulating the hysteresis loss in the audio transformers? And if so, how did you keep the CPU load so low?
This is a delicate question, as we do not want to disclose a confidential part of our technologies. But yes, we can go down to the level of component-level simulation if required. This makes it possible for us to apply both precise macro-behaviour modelling and component-level approximation techniques to one model. And it is true that we have established the modelling technology of audio transformers, which focuses on the exact realisation of the behaviour and the response of audio transformers. As for CPU load, we choose the appropriate modelling methods extremely carefully, in order to obtain the required sound quality while reducing the load.
What do you think is the key to an accurate emulation?
It depends on what's being modelled. In the case of phasers for the Motif XS synthesizer, CP1 electronic piano and the consoles, for example, the exact modelling, which means the exact digital transfer function mathematically, was the key. Once it was established, our digital phasers sounded very natural. When it comes to studio gear like compressors and EQs, it's not only the modelling but also fine-tuning for the right sound that is significant. Sometimes we spend more time tweaking the models and parameters than actually modelling the hardware.
Often, oversampling is used to overcome aliasing when generating harmonic overtones and avoid folding of the filter curve when approaching the Nyquist limit, but it's CPU-intensive and can add latency. How did you avoid using it with these plug-ins?
In many ways, oversampling's just a poor, inefficient means to realise models in the digital domain. For many applications, we can employ more mathematical methods to obtain the right filter responses. However, in some cases, oversampling is an essential technology, indispensable to reduce alias noises as you described. Fortunately, we avoided the oversampling strategy for the Portico EQ and compressor, making them free from undesirable latency problems.
What exciting news can we expect from the R&D teams of Yamaha and Steinberg in the future?
We are planning to release three more packages of VCM [Virtual Circuitry Modelling] audio effects. Many effects were initially developed for the digital consoles and others are from the Motif XS. They all provide a great, musical sound quality originating from analogue vintage gear. These plug-in effects are bound to emphasise the performance of Cubase and Nuendo. We are also targeting the live sound industry with the VCM technology. As you may know, many of our digital mixers are already equipped with VCM, and many people are using VCM as their favourite effects for live production. We will continue to use this technology for future products, as we think it is one of the greatest features of Yamaha consoles enjoyed by many of their users.